Knowledge Management in the Field

Start with the definitions, because they are load-bearing. Army doctrine (ATP 6-01.1, Techniques for Effective Knowledge Management) defines KM as "the process of enabling knowledge flow to enhance shared understanding, learning, and decision making." The stated objective is a five-part chain: "Getting the right information, to the right people, in the right format, at the right time, in order to arrive at the right decision." KM has four components that must be aligned and balanced — people (those who create, organize, apply, and transfer knowledge, and their leaders), process (the methods of creating, organizing, applying, sharing, and transferring it), tools (digital and non-digital), and organization (the matrix in which the other three function). And KM is executed as a five-step process: assess information flow, design the program, develop the solution, pilot it on a small scale with soldiers, implement it with training and monitoring. The primer's doctrinal anchor is heavier still: conducting KM is one of ADP 6-0's four primary staff tasks, and knowledge management is named as one of the five components of the mission command system itself, alongside people, the operations process, command posts, and the Army's information network. This is the direct continuation of session 05 — decision rights were delegated on the promise of shared understanding, and KM is the machinery that is supposed to produce it.

That foreword is the whole session in miniature: a mature, doctrinally mandated KM program, embedded at all levels for a decade — and field observations showing it routinely fails in execution. The primer names three recurring failure areas from the Combat Training Centers and the Mission Command Training Program: units cannot develop and maintain an effective common operational picture, cannot understand critical paths and battle rhythm, and cannot reliably access information across the Army. We return to all three in section 03 with the evidence.

The primer paraphrases doctrine, but we can read the doctrine itself, and read it across time. The Army rewrote its KM field manual at least three times in a decade — FM 6-01.1, Knowledge Management Operations (2012), then ATP 6-01.1, Techniques for Effective Knowledge Management (2015), then ATP 6-01.1, Knowledge Management (2024) — which gives us a rare thing: a controlled comparison of an institution editing its own definition of knowledge over nine years. The result is striking for how little the core moved. All three editions define KM identically as "the process of enabling knowledge flow to enhance shared understanding, learning, and decision making," carry the same four components (people, processes, tools, organization) and the same five-step process (assess, design, develop, pilot, implement), and repeat the "right information, right people, right format, right time, right decision" objective. The 2012 predecessor already framed KM as that five-step process. The definitional spine is frozen; what changed is the framing around it.

The 2024 edition is unusually candid about its own edits, and the candor is the point. Its introduction states plainly what changed — and every addition is a topic the 2015 doctrine's own field record showed it needed:

Read against the primer, this is the same tension one edition deeper. The definition insists KM is a human process; the 2024 additions — change management, sustaining a program, a validated maturity model — are all attempts to make that human process survive without a heroic commander, exactly the equilibrium problem the primer could not solve. Whether adding doctrine can close a gap that is fundamentally about executing doctrine is left open here; it is discussion question three by another name.

The primer's epistemology is a four-layer ladder — data (unprocessed, raw) → information (processed; structured, organized) → knowledge (synthesized; contextual, actionable) → shared understanding (team learning; insightful) — annotated with a claim about decision quality: as an organization climbs the ladder, decision behavior shifts from reactionary (high risk) to anticipatory (low risk). "There is a direct correlation between decision behavior and risk to shared understanding," the primer states. Seminar veterans will recognize this as the same cognitive hierarchy NDP 6 built in session 04 — the two services codify an identical DIKW-style ladder — but the primer adds the organizational move: the top rung is not individual understanding but shared understanding, produced by team learning. That is Wegner's move too: cognition as a group-level property.

Finally, note who the primer holds responsible. Not the technologists: "Knowledge management is primarily a human process and is heavily dependent on organizational leadership for direction and supervision." The senior leader must function as the organization's "knowledge leader" — understanding KM doctrine, establishing the program, providing knowledge requirements, implementing training, focusing effort "on knowledge processes instead of knowledge tools," and personally using and evaluating the products. The primer is blunt that identifying the commander's own knowledge requirements — a KM reading of the commander's critical information requirements — "often gets lost in the rush to implement technical tools." An organizational theorist should find this simultaneously impressive and alarming: impressive because the doctrine assigns knowledge work a named owner at the top; alarming because a system that only works when the commander personally drives it has no equilibrium behavior — a design question we take up in discussion.

Wegner (1987) described transactive memory as a group-level memory system built from individual memories plus a shared directory: members specialize, know who holds what expertise, trust each other's domains, and coordinate retrieval. Lewis (2003) measures it by three indicators — specialization, credibility, and coordination. The Army's KM team, specified in the primer's Appendix A, is this structure made explicit. The commander is decision maker and knowledge leader, supplying guidance and intent. The chief of staff or executive officer must "visualize the staff's processes for collecting, analyzing, and transferring knowledge," and ensure the KMO "has visibility of all staff processes and a means to implement change in those processes." The knowledge management officer directs the KM section and functions as a "knowledge catalyst" — the primer insists the successful KMO "must be well grounded in mission command, the operations process, and the military decisionmaking process," while an understanding of technology is "useful, but not critical." Knowledge management representatives sit inside each staff section as developers and "enforcers" of metadata standards and "managers of knowledge flow." The knowledge management working group, chartered by the chief of staff as an extension of the staff, is "the primary venue for the leadership to assess and improve knowledge flow within the organization."

Read through Lewis, "the SharePoint guy" is a credibility failure, not a job-description dispute. A transactive memory system only functions if members trust the directory — if the staff believes the KMO's map of who-knows-what and follows the retrieval paths it prescribes. A KMO coded as an IT administrator has no credibility over the operations process, so the directory he maintains goes unused, and each staff section falls back on its private networks. The deeper theoretical problem is that Wegner's and Lewis's TMS develops through shared experience — Liang, Moreland & Argote's (1995) classic finding is that groups trained together outperform groups of individually trained members precisely because training together builds the directory. Army units cannot rely on that: personnel rotate constantly (the primer notes every unit "especially in the early stages of sustainable readiness, experiences problems with untrained or incompletely trained Soldiers"). So the institution writes the directory down — battle rosters, terms-of-reference documents ("a summary of key people within an organization and their specific roles and responsibilities," which "enhances shared understanding regarding roles and responsibilities"), continuity files, an expertise location system. The KM program is an attempt to make transactive memory survive turnover by codifying the transactive layer itself.

Does it work? The primer's own evidence, drawn from Combat Training Center and MCTP observations, says: often not. The three vignettes are worth close reading because each is a different classical failure.

The primer's answer to "how do we even know where we stand" is the Knowledge Management Maturity Model (KM3) — developed from a 2009 Army War College paper, drawing on Carnegie Mellon's and APQC's private-sector maturity models. It defines 42 KM efforts grouped under the four components (people, process, tools, organization/culture), each scored on a 1–5 Likert scale, rolled up into five maturity levels: ad hoc (chaotic), repeatable, defined, managed, optimized. The primer's worked example scores 119 of 210 points — 56.7 percent overall maturity — with the telling detail that tools and culture outscore process. Even the illustrative unit is better at buying tools than at running knowledge processes.

One more institutional detail completes the picture: KM is a credentialed specialty. The Army Knowledge Management Qualification Course awards a Professional Skill Identifier (SI/ASI 1E), and a distance-learning basics course sits on the Army Learning Management System. The institution has professionalized the role of directory-keeper. Whether a transactive memory system can be staffed into existence — rather than grown through shared experience, as the theory says it must — is discussion question two.

All four maps agree on the skeleton. The Navy Warfare Library (NWL) — governed by its own "manual for manuals," NTRP 1-01 — organizes Navy publications into product lines by function and altitude: NDPs (capstone philosophy; six were conceived in the 1990s, and the maps agree NDP 3, Naval Operations, was never published — the capstone series has a permanent hole), NWPs (keystone operational doctrine, numbered to mirror the joint J-code series: 1 general, 2 intelligence, 3 operations, 4 logistics, 5 planning, 6 C4), NTTPs (tactical how-to for platforms and mission areas), NTRPs (reference data — tables, planning factors, terminology), FXPs (fleet exercise publications), and the agile layer of TACMEMOs, TACBULs, and TACNOTEs. Custody of the whole is federated: OPNAV N7 sets policy; the Navy Warfare Development Center (NWDC) in Norfolk is executive agent, publisher, and approval authority; the Warfighting Development Centers — SMWDC for surface and mine warfare, NAWDC (home of TOPGUN) for aviation, UWDC for undersea, NIWDC for information warfare — serve as Primary Review Authorities so that "a manual on missile defense is written by the experts … who practice it daily, rather than by a generalist staff officer"; fleet and type commanders act as Coordinating Review Authorities who verify the doctrine is physically executable.

The TACMEMO deserves a theorist's attention. It is a time-boxed exploration channel in March's (1991) sense: a provisional tactic pushed to the fleet for evaluation, carrying an evaluation plan and a feedback obligation, that self-cancels at roughly 24 months unless validated into permanent doctrine. The design solves the exploration–exploitation problem structurally — experiments cannot silently harden into doctrine, and doctrine cannot silently block experiments — while leaving the slow core untouched. Whether that is organizational ambidexterity or a pressure-release valve that lets the core stay slow is discussion question four.

The lifecycle around the library (per the Architecture of Naval Wisdom map, citing OPNAVINST 3510.17) runs in five phases: assessment and initiation (a validated need produces a Program Directive assigning the PRA and milestones); development and drafting (SME working groups — the map reports that replacing emailed Word drafts with a web-based collaborative review tool cut review timelines "by up to 50%, reducing a 6-week process to under 3 weeks"); critical review by cross-domain CRAs plus legal and security review; approval and promulgation (a NAVPUB general message, upload to the library); and maintenance (urgent changes by message for safety-critical errors; a mandatory periodic review — typically five years — after which an unreviewed publication risks automatic cancellation). Distribution is its own infrastructure: the Navy Doctrine Library System (NDLS), a CAC-gated SharePoint-hosted portal on both unclassified and classified networks, and — because ships fight under emissions control with no connectivity — Collaboration at Sea, which replicates the library onto shipboard servers so "a destroyer operating in the middle of the Pacific, under strict radio silence, still has access to the latest tactical guidance." The Army's parallel machinery: the Army Publishing Directorate authenticates, the Central Army Registry serves as "single source of truth," EPUB formats make manuals reflowable on phones, and doctrine now ships as apps and even audiobooks.

The bottom lanes matter as much as the top. The conversation-export map is the only one that traces knowledge all the way down: doctrine becomes human behavior through NAVEDTRA curricula and Personnel Qualification Standards — "a written compilation of knowledge and skills required to qualify for a watch station" — each PQS maintained by a designated Model Manager command; and it becomes machine behavior through NSTM chapters and the Engineering Operational Sequencing System, whose casualty-control procedures are formalized down to document type definitions. Knowledge flow in this institution is not complete when a publication is promulgated; it is complete when a qualification standard, a curriculum, and a step-by-step plant procedure have all been rewritten to match. In Hansen's terms this is codification pursued to its logical extreme — people-to-documents at every altitude from warfighting philosophy to pump maintenance, with reuse economics to match (write once at NWDC, execute on every ship).

The intake ritual is the After Action Review, which the Cognitive Citadel map calls "arguably the most significant knowledge management invention of the modern US Army." We can retire the hedge in that claim by going to the source. The AAR is not merely the four diagnostic questions everyone remembers — What was supposed to happen? What actually happened? Why did it happen? What will we do next time? — governed by one norm, that it is blameless, "focused on learning rather than discipline," so honesty is cheap. Underneath the questions is a four-step process, specified since 1993 in TC 25-20, A Leader's Guide to After-Action Reviews, that turns a debrief into a repeatable organizational routine with resource costs and a feedback obligation.

The four steps matter more than the four questions, because they are where the routine either learns or does not. TC 25-20 makes "following up" a step in its own right — "Fix the problem — retrain immediately, revise standing operating procedures … use to assist in making commander's assessment" — so a review that produces no changed behavior has skipped a doctrinal step, not merely underperformed. This is a lineage worth naming: TC 25-20 (1993) is the origin; the Army Research Institute's Foundations of the After Action Review Process (Special Report 42, 1999) is the empirical study that put the ritual on a research footing; and FM 7-0, Training (2021) is its current doctrinal home, treating the AAR in the execution chapter and Appendix K. A thirty-year-old feedback loop, continuously re-validated — which is precisely why the maps can call it the Army's signature KM invention. The Navy's tactical equivalent embeds intake in the Plan–Brief–Execute–Debrief cycle, where the debrief grades execution against the doctrinal standard and asks the diagnostic question that makes the whole system self-correcting: was the deviation operator error, or is the doctrine itself flawed? Every training event thereby becomes a data point with a routing rule attached.

Above the ritual sits the machinery. On the Army side, the Center for Army Lessons Learned (Fort Leavenworth, since 1985 — the publisher of this session's primer) collects AARs, embeds collection teams with deployed units, analyzes trends across thousands of reports, disseminates the "beta patch" products, and hands validated lessons to the Combined Arms Doctrine Directorate for formal incorporation — closing the loop. On the Navy side, the Navy Lessons Learned Program, with NWDC as program director, runs the Navy Lessons Learned Information System — an enclave of the DoD-wide Joint Lessons Learned Information System, holding "over 90,000 documents" as of 2013, searchable by keyword, filterable by type, and — critically — tagged to the Universal Joint Task List and Universal Naval Task List, so a lesson is indexed to the doctrinal task it bears on. The process is a filter chain: units submit observation reports after deployments and exercises; analysts validate — the Architecture of Naval Wisdom map's example is filtering "my coffee maker broke" from "the logistics software fails in low-bandwidth environments"; validated lessons get a Remedial Action Project that drives change across DOTMLPF-P — doctrine, organization, training, materiel, leadership, personnel, facilities, policy — the checklist that forces every fix to consider non-materiel solutions before purchases and prevents "orphan" capabilities where the weapon arrives but the manual, training, and maintenance bays do not.

1. 01 Unit level
   Observe and record. AAR after every training and combat event; PBED debrief against the doctrinal standard; observation reports submitted to NLLIS/JLLIS after deployments and major exercises.
2. 02 CALL / NWDC analysts
   Validate and aggregate. Filter local noise from systemic signal; synthesize trends across thousands of reports; link lessons to UJTL/UNTL tasks and to the doctrine they implicate.
3. 03 Fast channel
   Publish interim guidance. CALL handbooks, bulletins, and Initial Impressions Reports; Navy TACMEMOs, bulletins, and fleet messages — faster than doctrine, explicitly provisional.
4. 04 Doctrine authorities
   Institutionalize. Change recommendations to NWPs/NTTPs or FMs/ATPs; Remedial Action Projects drive DOTMLPF-P changes so the fix arrives with its organization, training, and materiel.
5. 05 Training and technical base
   Instantiate in people and machines. Schoolhouse curricula updated (NAVEDTRA; the Army's ADDIE process — knowledge is not "transferred" until embedded in the program of instruction); PQS revised by Model Managers; NSTM/EOSS technical procedures aligned.

Now the theory. Cohen & Levinthal (1990) defined absorptive capacity as an organization's ability to recognize the value of new knowledge, assimilate it, and apply it — and argued it is a function of prior related knowledge: you can only absorb what you already almost know. The military's loop is a deliberate, expensive construction of absorptive capacity. The UJTL/UNTL tagging scheme gives every incoming lesson a slot in an existing knowledge structure — that is prior related knowledge, institutionalized as taxonomy. The validation stage is a recognition filter staffed by experts. And the most interesting investment is human: NWDC embeds lessons-learned personnel at each numbered fleet, who "push" relevant lessons to staffs during planning and "pull" lessons from the repository on request — boundary spanners compensating for the fact that a database, however large, cannot recognize relevance on a staff's behalf. The Navy's own governance shows it understands intake is not absorption: the annual Navy Lessons Learned Workshop convenes practitioners from every community — with Army CALL and Air Force counterparts invited to benchmark — and its 2022 consensus was that the process must be "flexible and agile to support different organizational missions … while providing common programmatic standards to share collective insights." The current cultural push, the CNO's "Get Real, Get Better" initiative, frames candid self-assessment as a command responsibility, with the lessons program as "the framework and tools required to facilitate the transfer of timely, relevant and credible information across the fleet and accelerate learning."

Put the clock speeds side by side. The Architecture of Naval Wisdom map states that "the acceleration of warfare, driven by unmanned systems, artificial intelligence, and cyber capabilities, threatens to outpace the standard 18-to-24-month doctrine revision cycle" (elsewhere it says three years; the mandatory review threshold is five — see the conflict register in section 04). The Cognitive Citadel sharpens it with Ukraine: "While Ukraine iterates drone tactics weekly (e.g., FPV drone swarms), US doctrine development involves multi-year review cycles. This creates a dangerous 'doctrine gap.'" The institutional responses are all attempts to shorten write-latency without sacrificing validation: TACMEMOs and Rapid Action Revisions; CALL handbooks as the fast channel; the Navy's Task Force 59 experimenting with unmanned vessels in live operations, generating tactics "much faster than the traditional PRA process"; Doctrine 2015's stratification, which deliberately isolated fast-changing techniques (ATPs) from stable principles (ADPs) so the two could update on different clocks; and "Agile doctrine" — releasing "Beta" versions to the fleet and iterating. The most structurally interesting proposal is Web-Enabled Doctrine: decomposing publications from static PDFs into XML modules so that "a change in a specific technical specification (e.g., the range of a missile) [can] automatically propagate across all relevant manuals" — modularity applied to the knowledge base itself, dependencies made explicit so updates localize.

Latency is only the first failure mode. The second is adoption. The Gemini map's critique is blunt enough to quote in full:

This is Hansen's codification economics failing at the last yard: the strategy's entire payoff is cheap reuse, and reuse is priced not at the repository but at the reader. A library the sailor cannot open from the barracks — no personal-device access, CAC-gated portals, security constraints commercial KM never faces — collects its returns only on duty, on a government terminal, with intent. The security trade-off may nonetheless be correct; whether KM theory built on consultancies can even price that trade-off is discussion question five. The third failure mode is the one Brown & Duguid (1991) predicted: canonical practice diverges from noncanonical practice. The Cognitive Citadel documents an "interpretation gap" between "Espoused Doctrine (what the book says) and Theory-in-Use (what units actually do)": mission command preaches decentralized trust while "the risk-averse nature of garrison bureaucracy often trains leaders to seek permission for everything," a system the map calls "schizophrenic" — "the manual says 'Go,' but the culture says 'Stop.'" The conversation-export map adds the temporal version: units adapt faster than manuals, so best practices "live in SOPs, briefings, or unofficial docs before they become formal doctrine." And where does knowledge live in the meantime? In communities of practice. The Gemini map is explicit that the Navy "relies heavily on the CPO community as the repository of tacit knowledge" — the Chiefs' Mess as retention bin — and that communities of practice ("Logistics Planners," "Helicopter Pilots") use collaborative tools "to bypass formal hierarchies and solve problems laterally." The Army's MilWiki experiment institutionalized exactly this and then discovered its limits: it "serves more as a collaborative feedback tool than a direct publishing platform for authoritative doctrine," because authority still requires validation.

1. The Hansen test. Hansen, Nohria & Tierney warn that firms which straddle codification and personalization fail, and that winners pick one at roughly 80–20. The military runs both at full intensity: a total codification stack (NWL to EOSS) and a total personalization apparatus (KMOs, embedded fleet reps, the Chiefs' Mess, working groups). Is it the exception that disproves the rule — because it separates the strategies by knowledge type, codifying the explicit and personalizing the tacit — or do the CTC failures and the unread repositories show Hansen collecting his fee?
2. Staffing a group mind. Lewis's transactive memory develops through shared experience; Liang, Moreland & Argote showed you cannot substitute individually trained members. The Army, facing constant rotation, tries to codify the transactive layer itself — battle rosters, terms of reference, a credentialed KMO with an ASI 1E identifier. Can a TMS be staffed and documented into existence, or does the "SharePoint guy" credibility failure show that the directory only works when it is grown?
3. The heroic-leader dependency. The primer's most repeated claim is that KM programs succeed only with active senior-leader involvement. Is a knowledge system that requires the commander's personal drive well-designed (attention from the top is the scarcest, therefore most credible, signal) or badly designed (it has no stable equilibrium — the program decays the day the committed commander rotates out)? What would a KM system that runs at equilibrium even look like in a two-year command-tour institution?
4. TACMEMO as ambidexterity. March's exploration–exploitation tension is usually managed structurally (separate units) or temporally (cycles). The TACMEMO manages it by expiry date: exploration that self-cancels in 24 months unless validated. Is this an elegant ambidexterity mechanism — or a safety valve that protects the slow core from the pressure that would otherwise force it to speed up? What evidence would distinguish the two?
5. Pricing the CAC wall. The "digital native" critique treats the CAC-locked, no-personal-device library as a UX failure. But the classified/unclassified boundary is a real constraint that Hansen's consultancies and Brown & Duguid's copier technicians never faced. Does classical KM theory have any way to price the security–absorption trade-off, or does the military case expose a boundary condition of the whole literature?
6. Whose memory is it? If the Chiefs' Mess is the real repository of tacit knowledge and communities of practice solve problems laterally around the formal system, what does the codified library actually do — transfer knowledge, or legitimate and standardize knowledge that was transferred by other means? (Recall Meyer & Rowan from session 02: formal structure as myth and ceremony.) What observation would tell you which is happening?
7. Taxonomy as absorptive capacity — and as lockout. Tagging every lesson to the Universal Joint Task List gives incoming knowledge a slot in prior knowledge (Cohen & Levinthal's condition for absorption) and makes lessons searchable by doctrinal task. It also forces the genuinely novel into old categories. Where in the mapped system is the mechanism for absorbing a lesson that the task list cannot classify — and if you cannot find one, what does that predict?
8. Studying memory through unreliable memory. Our own maps of this system are AI syntheses that disagree on the system's basic clock speed and partially plagiarize each other. What does the experience of cross-checking them teach about provenance, validation, and authority as KM design requirements — the very goods the publication-centric model supplies and the knowledge-centric model risks?